The vast repertoire of immunoglobulins and T cell receptors is generated, in part, by V(D)J recombination, a series of genomic rearrangements that occur specifically in developing lymphocytes. The recombination activating gene, RAG-1, which is a gene expressed exclusively in maturing lymphoid cells, was previously isolated. RAG-1 inefficiently induced V(D)J recombinase activity when transfected into fibroblasts, but cotransfection with an adjacent gene, RAG-2, has resulted in at least a 1000-fold increase in the frequency of recombination. The 2.1-kilobase RAG-2 complementary DNA encodes a putative protein of 527 amino acids whose sequence is unrelated to that of RAG-1. Like RAG-1, RAG-2 is conserved between species that carry out V(D)J recombination, and its expression pattern correlates precisely with that of V(D)J recombinase activity. In addition to being located just 8 kilobases apart, these convergently transcribed genes are unusual in that most, if not all, of their coding and 3' untranslated sequences are contained in single exons. RAG-1 and RAG-2 might activate the expression of the V(D)J recombinase but, more likely, they directly participate in the recombination reaction.

The RAG-1 (recombination activating gene-1) genomic locus, which activates V(D)J recombination when introduced into NIH 3T3 fibroblasts, was isolated by serial genomic transfections of oligonucleotide-tagged DNA. A genomic walk spanning 55 kb yielded a RAG-1 genomic probe that detects a single 6.6-7.0 kb mRNA species in transfectants and pre-B and pre-T cells. RAG-1 genomic and cDNA clones were biologically active when introduced into NIH 3T3 cells. Nucleotide sequencing of human and mouse RAG-1 cDNA clones predicts 119 kd proteins of 1043 and 1040 amino acids, respectively, with 90% sequence identity. RAG-1 has been conserved between species that carry out V(D)J recombination, and its pattern of expression correlates exactly with the pattern of expression of V(D)J recombinase activity. RAG-1 may activate V(D)J recombination indirectly, or it may encode the V(D)J recombinase itself.

Formation of double-strand breaks at recombination signal sequences is an early step in V(D)J recombination. Here we show that purified RAG1 and RAG2 proteins are sufficient to carry out this reaction. The cleavage reaction can be divided into two distinct steps. First, a nick is introduced at the 5' end of the signal sequence. The other strand is then broken, resulting in a hairpin structure at the coding end and a blunt, 5'-phosphorylated signal end. The hairpin is made as a direct consequence of the cleavage mechanism. Nicking and hairpin formation each require the presence of a signal sequence and both RAG proteins.

Nuclear processes such as transcription, DNA replication, and recombination are dynamically regulated by chromatin structure. Transcription is known to be regulated by chromatin-associated proteins containing conserved protein domains that specifically recognize distinct covalent posttranslational modifications on histones. However, it has been unclear whether similar mechanisms are involved in mammalian DNA recombination. Here, we show that RAG2 -an essential component of the RAG1/2 V(D)J recombinase, that mediates antigen receptor gene assembly 1 -contains a plant homeodomain (PHD) finger that specifically recognizes histone H3 trimethylated at lysine 4 (H3K4me3). The high-resolution crystal structure of the RAG2 PHD finger bound to H3K4me3 reveals the molecular basis of H3K4me3-recognition by RAG2. Mutations that abrogate RAG2's recognition of H3K4me3 severely impair V(D)J recombination in vivo. Reducing the level of H3K4me3 similarly leads to a decrease in V(D)J recombination in vivo. Notably, a conserved tryptophan residue (W453) that constitutes a key structural component of the K4me3-binding surface and is essential for RAG2's recognition of H3K4me3 is mutated in patients with immunodeficiency syndromes. Together our results identify a novel function for histone methylation in mammalian DNA recombination. Furthermore, our results provide the first evidence suggesting that disrupting the read-out of histone modifications can cause an inherited human disease. +To whom correspondence should be addressed: oettinger@frodo.mgh.harvard.edu; ogozani@stanford.edu. * These authors contributed equally to the work Note added in proof: While this work was under review, another study also reported that the RAG2 PHD finger binds to methylated H3K4 30 .Atomic coordinates and structure factors of the RAG2 PHD -H3K4me3 peptide complex have been deposited in the Protein Data Bank with the accession code of 2v89. Reprints and permissions information is available at npg.nature.com/reprintsandpermissions.Supplementary Information is linked to the online version of the paper at www.nature.com/nature. Since RAG2 contains a noncanonical plant homeodomain (PHD) finger 6,7 -a module that can mediate interactions with chromatin 8-10 -we asked whether a polypeptide encompassing the RAG2 PHD finger (RAG2 PHD : aa 414-527) can recognize modified histone proteins. In an in vitro screen of peptide microarrays containing ~70 distinct modified histone peptides, we found that RAG2 PHD specifically binds to histone H3 trimethylated at lysine 4 (H3K4me3) ( Fig. 1a ; Fig. S1; data not shown). The specificity of this interaction was confirmed by peptide pulldown assays ( Fig. 1b ; Fig. S2; Fig. S3). RAG2 has a C-terminal extension of 40 aa that is essential for phosphoinositide (PtdInsP)-binding 7 (aa 488-527), but this region is dispensable for H3K4me3-binding as the minimal PHD finger alone (aa 414-487) is sufficient for H3K4me3-recognition (Fig. 1c). In addition, the acidic hinge region of RAG2 (aa 388-412), previously implicated in...